[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7DM6JAEF7EC25HW02KSQY40":3,"skill-same:01M7DM6JAEF7EC25HW02KSQY40":56},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":24,"commitSha":25,"skillDirectory":26,"forksCount":27,"repositoryHtmlUrl":28,"repositoryHomepageUrl":6,"ownerHtmlUrl":29,"ownerBlogUrl":30,"ownerDescription":6,"updatedAt":31,"install":32,"resources":45},"01M7DM6JAEF7EC25HW02KSQY40","runtime-slice-design",null,"Designs buildable slices of the Master FinHub Python runtime (agent loop, context compaction, safety\u002Fsandbox, router, DAG engine, message bus, docker sandbox, MCP client, SSE server, evals) or the adoption of one backlog capability into the finhub-harness plugin, agents or runtime, from the reference port maps; each design has files, typed interfaces or insertion anchors, a proof, a test plan with mutation targets, and an Authority List mapping every claim to references\u002F\u003Csubmodule>\u002F\u003Cpath>:\u003Cline>. Use for: design the slices, slice plan, redesign slice N, design the adoption of C3, revise rejected claims, update the design after I changed the spec. Also read by runtime-builder before building. Not for general Python architecture outside this repo or for designing MCP servers (Mercury, Twilio or any other).","danielnguyenfinhub\u002Ffinhub-harness","danielnguyenfinhub","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F264356854?v=4","01M7DM55XETDR9HF92FWC28SBN","Daniel Nguyen","01M7DM55YGJYBWW488JZSQYEJ4",1,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Fdanielnguyenfinhub\u002Ffinhub-harness\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fruntime-slice-design\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: runtime-slice-design\ndescription: \"Designs buildable slices of the Master FinHub Python runtime (agent loop, context compaction, safety\u002Fsandbox, router, DAG engine, message bus, docker sandbox, MCP client, SSE server, evals) or the adoption of one backlog capability into the finhub-harness plugin, agents or runtime, from the reference port maps; each design has files, typed interfaces or insertion anchors, a proof, a test plan with mutation targets, and an Authority List mapping every claim to references\u002F\u003Csubmodule>\u002F\u003Cpath>:\u003Cline>. Use for: design the slices, slice plan, redesign slice N, design the adoption of C3, revise rejected claims, update the design after I changed the spec. Also read by runtime-builder before building. Not for general Python architecture outside this repo or for designing MCP servers (Mercury, Twilio or any other).\"\n---\n\n# Runtime Slice Design\n\nTurn the fourteen port maps into an ordered set of small, independently provable slices, or one backlog row into one provable adoption. Each slice must be buildable in one builder call and verifiable by four commands; anything bigger hides failures.\n\n## Steps\n\n1. **Read every port map** (`_workspace\u002F01_reference-miner_*_portmap.md`). Note which have `Status: PARTIAL` or empty Findings — claims depending on them must be marked, not invented.\n2. **Follow the slice order** in `.claude\u002Fskills\u002Fmaster-finhub-orchestrator\u002Freferences\u002Fphase-table.md` (Part B). Do not reorder: later slices assume earlier interfaces (the router in 4 plugs into the loop from 1; the DAG in 6 runs the checkpoint from 5).\n3. **Write each slice** using `references\u002Fslice-template.md`: files, interfaces with Python type signatures, proof test, deps, ported-vs-net-new, quant guardrails if relevant, and the cost statement for any refusal, prompt or required step the slice adds (`skills\u002Ffinhub-harness\u002Freferences\u002Fquality-gates.md` section 3-11; adapted from references\u002Fopenrig\u002FCONTRIBUTING.md:29-31 (Apache-2.0)).\n4. **Write the Authority List** at the end per `references\u002Fauthority-list.md`. Every design claim that relies on how a reference does something gets a row; net-new decisions get a `NET-NEW` row with a reason. Then run `python3 skills\u002Ffinhub-harness\u002Fscripts\u002Fcite_check.py \u003Cdesign file>` from the project root and fix every `BAD` line before step 5; on exit 3 name the submodules it could not read in the \"design ready\" message (`skills\u002Ffinhub-harness\u002Freferences\u002Fquality-gates.md` section 1-4).\n5. **Signal the judge** (\"design ready\") and wait. On a verdict with REJECTED > 0, fix only those claims and the slice text they support.\n\n## Design rules and why\n\n- **Python 3.12, stdlib first.** Every dependency is a supply-chain and licence question; slices 1–3, 5–7 and 11 need none. `graphlib.TopologicalSorter` over `graphon` for the DAG — dify pins an external package we would then have to audit.\n- **Protocols at seams.** The LLM is a `typing.Protocol` with a `FakeLLM` for tests, so no slice before 4 needs network or keys. Tools declare a schema the loop validates against — boundary-qa compares these shapes directly.\n- **Proof = one command or one test.** If you cannot state the proof in one line, the slice is too big; split it.\n- **Mypy-strict friendly.** Declare return types, avoid `Any` at public boundaries, use `dataclass(frozen=True)` for messages and checkpoints (immutable data is easier to checkpoint and resume).\n- **Safety slices are policy-driven and small.** Slice 3 and 8 are net-new where references lack them; keep each ≤80 lines and explain the denylist in the Authority List as `NET-NEW (spec requirement)`.\n- **Quant correctness is designed in.** Any slice touching evals, verifiers, backtests or pricing states how it handles each guardrail in `.claude\u002Fskills\u002Fadversarial-audit\u002Freferences\u002Fquant-guardrails.md`. The judge will check those rows regardless.\n- **Licences.** Never cite `references\u002Fautogpt\u002Fautogpt_platform\u002F`; dify claims are `pattern` only. Citations outside `references\u002F` are invalid.\n- **Synthetic data only.** Example payloads in the design use placeholder names (`client_a`, `acct_001`), never real people, accounts or Mercury records.\n\n## Output\n\n`_workspace\u002F02_strategy-architect_slices.md`:\n\n```markdown\n# Master FinHub runtime — slice design (revision 1)\n\n## Slice 1 — loop + echo tool + CLI\n...(slice-template sections)...\n\n## Slice 2 — context compaction\n...\n\n## Authority List\n| id | claim | evidence | slice |\n|---|---|---|---|\n| A1 | ... | references\u002F...:42 | 1 |\n```\n\n## Adoption design (one backlog item)\n\nWhen the prompt names a backlog row (`C3`) instead of a slice range, write `_workspace\u002F02_strategy-architect_\u003Citem-id>.md` with these sections, in this order: `## Source` (the backlog row and its port-map rows, re-opened), `## Target` (plugin skill \u002F agent \u002F runtime module \u002F docs, and the exact files), `## Design` (for code: files and typed interfaces; for prose: each insertion as anchor line quoted verbatim + the text to insert + the one-line attribution `adapted from references\u002F\u003Crepo>\u002F\u003Cpath>:\u003Cline> (\u003Clicence>)`), `## Proof` (the commands or greps that show every insertion or interface landed, with expected counts), `## Test plan` (unit cases, must-still-pass list, and at least three mutation targets QA can apply on a scratch copy, each with the one test id it should break and the text of the assertion it should trip (`test` and `sig`, `skills\u002Ffinhub-harness\u002Freferences\u002Fquality-gates.md` §3-4; required in an adoption's plan, optional in a runtime slice's, and the judge checks that each `sig` occurs in a test; adapted from references\u002Fopenrig\u002Fpackages\u002Ftest-system\u002Fci\u002Fresult.mjs:33-34 (Apache-2.0)), `## Does not cover` (honest limits and known false positives), then `## Cost statement` (one row per refusal, prompt or required step the adoption adds, with scenario, who causes it and cost to others, or the one line `Cost statement: none; this design adds no refusal, prompt or required step.`), then `## Authority List`. Keep existing tests unmodified; a design that needs an existing test changed says why in `## Design`. A prose adoption never pastes source prose: an 8-word run identical to the source is a defect. Before saying an adoption design is ready, run the cite check of step 4 on it and fix every `BAD` line (`skills\u002Ffinhub-harness\u002Freferences\u002Fquality-gates.md` section 1-4).\n\n## On revision\n\nRead the latest verdict for your design (`02_adversarial-risk-judge_verdict.md` for runtime slices; `02_adversarial-risk-judge_\u003Citem-id>_r\u003Ck>.md` with the highest `k` for an adoption). Change only REJECTED\u002FUNVERIFIED claims; keep UPHELD rows byte-identical so the judge can carry them forward. Bump the revision number in the title and list changed claim ids at the top under `## Changes in revision N`. If you believe a rejection is wrong, reply to the judge with the exact line as new evidence rather than resubmitting the same claim.\n\n## For the builder\n\nruntime-builder reads this skill to understand the slice template, then implements only UPHELD claims for its slice. Interfaces declared here are the contract boundary-qa checks against — if the builder must deviate, it records why in its report.\n","# Runtime Slice Design\n\nTurn the fourteen port maps into an ordered set of small, independently provable slices, or one backlog row into one provable adoption. Each slice must be buildable in one builder call and verifiable by four commands; anything bigger hides failures.\n\n## Steps\n\n1. **Read every port map** (`_workspace\u002F01_reference-miner_*_portmap.md`). Note which have `Status: PARTIAL` or empty Findings — claims depending on them must be marked, not invented.\n2. **Follow the slice order** in `.claude\u002Fskills\u002Fmaster-finhub-orchestrator\u002Freferences\u002Fphase-table.md` (Part B). Do not reorder: later slices assume earlier interfaces (the router in 4 plugs into the loop from 1; the DAG in 6 runs the checkpoint from 5).\n3. **Write each slice** using `references\u002Fslice-template.md`: files, interfaces with Python type signatures, proof test, deps, ported-vs-net-new, quant guardrails if relevant, and the cost statement for any refusal, prompt or required step the slice adds (`skills\u002Ffinhub-harness\u002Freferences\u002Fquality-gates.md` section 3-11; adapted from references\u002Fopenrig\u002FCONTRIBUTING.md:29-31 (Apache-2.0)).\n4. **Write the Authority List** at the end per `references\u002Fauthority-list.md`. Every design claim that relies on how a reference does something gets a row; net-new decisions get a `NET-NEW` row with a reason. Then run `python3 skills\u002Ffinhub-harness\u002Fscripts\u002Fcite_check.py \u003Cdesign file>` from the project root and fix every `BAD` line before step 5; on exit 3 name the submodules it could not read in the \"design ready\" message (`skills\u002Ffinhub-harness\u002Freferences\u002Fquality-gates.md` section 1-4).\n5. **Signal the judge** (\"design ready\") and wait. On a verdict with REJECTED > 0, fix only those claims and the slice text they support.\n\n## Design rules and why\n\n- **Python 3.12, stdlib first.** Every dependency is a supply-chain and licence question; slices 1–3, 5–7 and 11 need none. `graphlib.TopologicalSorter` over `graphon` for the DAG — dify pins an external package we would then have to audit.\n- **Protocols at seams.** The LLM is a `typing.Protocol` with a `FakeLLM` for tests, so no slice before 4 needs network or keys. Tools declare a schema the loop validates against — boundary-qa compares these shapes directly.\n- **Proof = one command or one test.** If you cannot state the proof in one line, the slice is too big; split it.\n- **Mypy-strict friendly.** Declare return types, avoid `Any` at public boundaries, use `dataclass(frozen=True)` for messages and checkpoints (immutable data is easier to checkpoint and resume).\n- **Safety slices are policy-driven and small.** Slice 3 and 8 are net-new where references lack them; keep each ≤80 lines and explain the denylist in the Authority List as `NET-NEW (spec requirement)`.\n- **Quant correctness is designed in.** Any slice touching evals, verifiers, backtests or pricing states how it handles each guardrail in `.claude\u002Fskills\u002Fadversarial-audit\u002Freferences\u002Fquant-guardrails.md`. The judge will check those rows regardless.\n- **Licences.** Never cite `references\u002Fautogpt\u002Fautogpt_platform\u002F`; dify claims are `pattern` only. Citations outside `references\u002F` are invalid.\n- **Synthetic data only.** Example payloads in the design use placeholder names (`client_a`, `acct_001`), never real people, accounts or Mercury records.\n\n## Output\n\n`_workspace\u002F02_strategy-architect_slices.md`:\n\n```markdown\n# Master FinHub runtime — slice design (revision 1)\n\n## Slice 1 — loop + echo tool + CLI\n...(slice-template sections)...\n\n## Slice 2 — context compaction\n...\n\n## Authority List\n| id | claim | evidence | slice |\n|---|---|---|---|\n| A1 | ... | references\u002F...:42 | 1 |\n```\n\n## Adoption design (one backlog item)\n\nWhen the prompt names a backlog row (`C3`) instead of a slice range, write `_workspace\u002F02_strategy-architect_\u003Citem-id>.md` with these sections, in this order: `## Source` (the backlog row and its port-map rows, re-opened), `## Target` (plugin skill \u002F agent \u002F runtime module \u002F docs, and the exact files), `## Design` (for code: files and typed interfaces; for prose: each insertion as anchor line quoted verbatim + the text to insert + the one-line attribution `adapted from references\u002F\u003Crepo>\u002F\u003Cpath>:\u003Cline> (\u003Clicence>)`), `## Proof` (the commands or greps that show every insertion or interface landed, with expected counts), `## Test plan` (unit cases, must-still-pass list, and at least three mutation targets QA can apply on a scratch copy, each with the one test id it should break and the text of the assertion it should trip (`test` and `sig`, `skills\u002Ffinhub-harness\u002Freferences\u002Fquality-gates.md` §3-4; required in an adoption's plan, optional in a runtime slice's, and the judge checks that each `sig` occurs in a test; adapted from references\u002Fopenrig\u002Fpackages\u002Ftest-system\u002Fci\u002Fresult.mjs:33-34 (Apache-2.0)), `## Does not cover` (honest limits and known false positives), then `## Cost statement` (one row per refusal, prompt or required step the adoption adds, with scenario, who causes it and cost to others, or the one line `Cost statement: none; this design adds no refusal, prompt or required step.`), then `## Authority List`. Keep existing tests unmodified; a design that needs an existing test changed says why in `## Design`. A prose adoption never pastes source prose: an 8-word run identical to the source is a defect. Before saying an adoption design is ready, run the cite check of step 4 on it and fix every `BAD` line (`skills\u002Ffinhub-harness\u002Freferences\u002Fquality-gates.md` section 1-4).\n\n## On revision\n\nRead the latest verdict for your design (`02_adversarial-risk-judge_verdict.md` for runtime slices; `02_adversarial-risk-judge_\u003Citem-id>_r\u003Ck>.md` with the highest `k` for an adoption). Change only REJECTED\u002FUNVERIFIED claims; keep UPHELD rows byte-identical so the judge can carry them forward. Bump the revision number in the title and list changed claim ids at the top under `## Changes in revision N`. If you believe a rejection is wrong, reply to the judge with the exact line as new evidence rather than resubmitting the same claim.\n\n## For the builder\n\nruntime-builder reads this skill to understand the slice template, then implements only UPHELD claims for its slice. Interfaces declared here are the contract boundary-qa checks against — if the builder must deviate, it records why in its report.\n",{"name":5,"description":7},"main","dda161ff9daf0d9bccbdba26e34bbee0438a18b0",".claude\u002Fskills\u002Fruntime-slice-design",0,"https:\u002F\u002Fgithub.com\u002Fdanielnguyenfinhub\u002Ffinhub-harness","https:\u002F\u002Fgithub.com\u002Fdanielnguyenfinhub","www.finhub.net.au","2026-11-12 17:07:32",{"source":8,"skillName":5,"commands":33,"prompt":43,"zipAvailable":44},[34,37,40],{"tool":35,"command":36},"npx","npx skills add danielnguyenfinhub\u002Ffinhub-harness --skill runtime-slice-design",{"tool":38,"command":39},"bunx","bunx skills add danielnguyenfinhub\u002Ffinhub-harness --skill runtime-slice-design",{"tool":41,"command":42},"pnpm","pnpm dlx skills add danielnguyenfinhub\u002Ffinhub-harness --skill runtime-slice-design","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fdanielnguyenfinhub\u002Ffinhub-harness\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fruntime-slice-design\u002FSKILL.md",true,[46,52],{"path":47,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":50,"sourceHtmlUrl":51},"references\u002Fauthority-list.md","REFERENCE","参考文档",3253,"https:\u002F\u002Fgithub.com\u002Fdanielnguyenfinhub\u002Ffinhub-harness\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fruntime-slice-design\u002Freferences\u002Fauthority-list.md",{"path":53,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":54,"sourceHtmlUrl":55},"references\u002Fslice-template.md",3245,"https:\u002F\u002Fgithub.com\u002Fdanielnguyenfinhub\u002Ffinhub-harness\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fruntime-slice-design\u002Freferences\u002Fslice-template.md",[57,64,71,77,83,90],{"skillNo":58,"name":59,"nameZh":6,"description":60,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":61,"categories":62,"keywords":63,"publishedAt":19},"01M7DMBP1MQ9X39BGJ9VF62M9P","linky","Fixture whose links and bundled-file mentions all resolve. Use to prove no false errors.","https:\u002F\u002Fgithub.com\u002Fdanielnguyenfinhub\u002Ffinhub-harness\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fharness_links_ok\u002Fskills\u002Flinky\u002FSKILL.md",[],[],{"skillNo":65,"name":66,"nameZh":6,"description":67,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":68,"categories":69,"keywords":70,"publishedAt":19},"01M7DMBK70KXF6A3V1VDRHYGVT","dup-one","Second owner of the name. Use for the duplicate fixture.","https:\u002F\u002Fgithub.com\u002Fdanielnguyenfinhub\u002Ffinhub-harness\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fharness_links_bad\u002Fskills\u002Fdup-two\u002FSKILL.md",[],[],{"skillNo":72,"name":66,"nameZh":6,"description":73,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":74,"categories":75,"keywords":76,"publishedAt":19},"01M7DMBFWK8HD4AWNK89KGE89P","First owner of the name. Use for the duplicate fixture.","https:\u002F\u002Fgithub.com\u002Fdanielnguyenfinhub\u002Ffinhub-harness\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fharness_links_bad\u002Fskills\u002Fdup-one\u002FSKILL.md",[],[],{"skillNo":78,"name":79,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":80,"categories":81,"keywords":82,"publishedAt":19},"01M7DMBETX7172TFNBAA3BVREQ","nested","https:\u002F\u002Fgithub.com\u002Fdanielnguyenfinhub\u002Ffinhub-harness\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fharness_links_bad\u002Fskills\u002Fbroken\u002Fnested\u002FSKILL.md",[],[],{"skillNo":84,"name":85,"nameZh":6,"description":86,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":87,"categories":88,"keywords":89,"publishedAt":19},"01M7DMBDCH5G61QVAB1QFMQA55","broken","Fixture with one defect per rule. Use to prove each rule fires.","https:\u002F\u002Fgithub.com\u002Fdanielnguyenfinhub\u002Ffinhub-harness\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fharness_links_bad\u002Fskills\u002Fbroken\u002FSKILL.md",[],[],{"skillNo":91,"name":92,"nameZh":6,"description":93,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":94,"categories":95,"keywords":96,"publishedAt":19},"01M7DMBAQHY0Z2PTYGCPM8H7MT","notes-style","House style for client notes. Use when writing or editing a client note.","https:\u002F\u002Fgithub.com\u002Fdanielnguyenfinhub\u002Ffinhub-harness\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fharness_good\u002Fskills\u002Fnotes-style\u002FSKILL.md",[],[]]